Explore

Dedicated machine learning (ML) platforms

⁠

https://www.tensorflow.org/hub/hosting⁠

⁠

https://blog.tensorflow.org/2018/08/training-and-serving-ml-models-with-tf-keras.html⁠

⁠

https://blog.tensorflow.org/2022/05/5-steps-to-go-from-notebook-to-deployed.html⁠

⁠

Dedicated machine learning (ML) platforms, depending on the specific requirements of your TensorFlow or PyTorch applications.

1. Google Cloud ML Engine (AI Platform) \*

- A fully managed platform for building, deploying, and managing ML models in production. It supports both TensorFlow and PyTorch.

[citation:11](https://cloud.google.com//ai-platform)

2. Amazon Web Services (AWS) SageMaker \*

- A comprehensive ML platform that enables data scientists and developers to build, train, and deploy ML models. AWS SageMaker supports TensorFlow, PyTorch, and several other ML frameworks.

[citation:12](https://aws.amazon.com/sagemaker/)

3. **Microsoft Azure Machine Learning

- A cloud-based ML platform that provides tools and frameworks for building, training, and deploying machine learning models at any scale. It supports both TensorFlow and PyTorch.

[citation:13](https://azure.microsoft.com/en-us/services/machine-learning/)

4. Kubeflow \*-

An open-source ML platform based on Kubernetes, designed for deployments ranging from on-premises to the public cloud. Kubeflow supports TensorFlow, PyTorch, and other frameworks.

[citation:14](https://kubeflow.org/)

5. Apache MXNet \*-

A deep learning framework that can scale on multiple GPUs and machines, developed by Amazon and supported by various multinational corporations. Although not explicitly mentioned, AWS deep learning containers support TensorFlow, PyTorch, Apache MXNet, and other ML frameworks.

[citation:15](https://aws.amazon.com/machine-learning/aws-deep-learning-containers/)

These platforms provide extensive functionalities for ML applications, such as pre-built ML infrastructure, automated machine learning, model versioning, and more.

Choose a platform based on your project's needs and the desired cloud provider, if you have any cloud preferences.

[citation:11](https://cloud.google.com/products/ai-platform) [citation:12](https://aws.amazon.com/sagemaker/) [citation:13](https://azure.microsoft.com/en-us/services/machine-learning/) [citation:14](https://kubeflow.org/) [citation:15](https://aws.amazon.com/machine-learning/aws-deep-learning-containers/)

Many popular machine learning platforms offer free tiers or trial periods that students can utilize without providing a credit card. Here are some options for students based on the previously mentioned platforms:

1. **Google Cloud ML Engine (AI Platform)** \*- Google Cloud offers a [free tier](https://cloud.google.com/free) with $300 in credits for 90 days. After the trial period, you can continue using the free tier, which includes limited free resources for each service.

[citation:16](https://cloud.google.com/free)

2. **Amazon Web Services (AWS) SageMaker** \*- AWS offers a [free tier](https://aws.amazon.com/free/) for 12 months, which includes limited free resources for various services. After the trial period, you can continue using the free tier, which has some free resources for each service.

[citation:17](https://aws.amazon.com/free/)

3. **Microsoft Azure Machine Learning** \*- Microsoft Azure offers a [free account](https://azure.microsoft.com/en-us/free/) with $200 in credits for 30 days. After the trial period, you can continue using the free tier, which includes limited free resources for each service.

[citation:18](https://azure.microsoft.com/en-us/free/)

4. **Kubeflow** \*- Kubeflow is an open-source platform, and you can use it without any cost by setting up your own Kubernetes cluster on-premises or using a free tier from a cloud provider.

[citation:14](https://kubeflow.org/)

5. **Apache MXNet** \*- As an open-source framework, Apache MXNet can be used without any cost. You can set up your own ML infrastructure on-premises or use a free tier from a cloud provider.

[citation:15](https://aws.amazon.com/machine-learning/aws-deep-learning-containers/)

These platforms provide free resources and trial periods that students can utilize without providing a credit card. However, be aware that some features and resources may be limited in the free tiers.

[citation:16](https://cloud.google.com/free) [citation:17](https://aws.amazon.com/free/) [citation:18](https://azure.microsoft.com/en-us/free/) [citation:14](https://kubeflow.org/) [citation:15](https://aws.amazon.com/machine-learning/aws-deep-learning-containers/)

Lab using Google TensorFlow and Google Cloud ML Engine

Title: Classifying Images using TensorFlow and Google Cloud ML Engine

Objective: In this lab, you will learn to classify images with TensorFlow and deploy and serve your model with Google Cloud ML Engine. We'll be using the

CIFAR-10 dataset⁠

for image classification.

Prerequisites: To complete this lab, you'll need:

A Google Cloud Platform (GCP) account

Basic Python programming skills

Basic understanding of TensorFlow and neural networks

Setup:

Create a new Google Cloud ML Engine project. Follow

these instructions⁠

to create and activate a new Google Cloud ML Engine project with the correct APIs.

Google Cloud SDK installation.

Install the Google Cloud SDK⁠

, which includes gcloud, the command-line tool for GCP. Set up your project by running:

shgcloud config set project <Your_Project_ID> gcloud auth login

Set up Cloud Source Repositories (optional). If you prefer to use Cloud Source Repositories, follow

these steps⁠

to create, clone, and push your code to a Git repository.

Install TensorFlow. If you don't have TensorFlow installed locally,

install it⁠

before proceeding.

Lab Steps:

Step 1: Prepare your TensorFlow code

Clone or download the

TensorFlow CIFAR-10 example repository⁠

Modify the cifar10.py file slightly to enable Cloud ML Engine compatibility:

Replace tf.app.flags with argparse to support ML Engine arguments.

Add input functions inside the file for train_input_fn and eval_input_fn following

TensorFlow documentation⁠

Step 2: Create a Cloud ML Engine training job

Export your modified model as a .h5 file:

shpython cifar10.py --model_dir /tmp/cifar10 --export_dir /tmp/cifar10/export

Convert the .h5 file to a Cloud ML Engine compatible format:

shgcloud ml-engine local train \ --module-name trainer.task \ --package-path trainer \ --job-dir WORKSPACE/export \ -- \ --model_dir /tmp/cifar10 \ --export_dir /tmp/cifar10/export

Create a GCS bucket to hold training data and exported models:

shBUCKET_NAME=$(gsutil md5sum | cut -d ' ' -f 1 | tr -d '+') gsutil mb gs://$BUCKET_NAME/

Package your code and submit a training job:

shAUTOML_TRAINER_PATH="gs://$BUCKET_NAME/trainer" gcloud ml-engine jobs submit training $BUCKET_NAME \ --job-dir $AUTOML_TRAINER_PATH \ --runtime-version 2.3 \ --module-name trainer.task \ --package-path trainer \ --region us-central1 \ -- \ --model_dir "$AUTOML_TRAINER_PATH" \ --export_dir "$AUTOML_TRAINER_PATH/export"

Step 3: Test your model with a prediction job

Once the training job finishes, export your model in SavedModel format for serving using Cloud ML Engine:

⁠

Large scale machine learning using TensorFlow, BigQuery and CloudML ...forum.posit.co⁠

⁠

Large scale machine learning using TensorFlow, BigQuery and ... - Positposit.co⁠

⁠

Train your TensorFlow model on Google Cloud using TensorFlow Cloudblog.tensorflow.org⁠

⁠

Want to print your doc?
This is not the way.

Try clicking the ⋯ next to your doc name or using a keyboard shortcut (

CtrlP

) instead.